rf coherent massive mimo a better ran alternative for 4g
TRANSCRIPT
RF Coherent Massive MIMO A Better RAN Alternative for 4G and 5G
Mihai Banu, CTO, Blue Danube Systems
July 5, 2019 | Blue Danube Systems | EXTERNAL | 1
Outline
1. Brief review of conventional Digital Massive MIMO
2. Fundamental flaws of Digital Massive MIMO
3. Array RF coherency and True 3D Beams to the rescue
4. Blue Danube economical Coherent Massive MIMO
5. MIMO technology evolution map and conclusions
July 5, 2019 | Blue Danube Systems | EXTERNAL | 2
Generic MIMO System Architecture
• “Digital” Radio Chains❑ ADC/DAC pairs + supporting analog/RF/digital circuitry
❑ Each radio chain has at least one RF Stage
❑ Each RF Stage connected to at least one antenna
• Ports: Independent data pipes between Processor and Radio Chains
• Layers: Independent data streams between Base Station and UEs
• L ≤ channel degrees of freedom
• In general: L ≤ P ≤ N ≤ M ≤ T (very important for minimizing HW)
• SU-MIMO (layers to one UE) & MU-MIMO (layers to multiple UEs)
T
Antennas
L
LayersBaseband
Processor
N
ADC/DAC
Stages
M
RF
StagesP Ports
…
…
… …
UE1
UE2
UE3
Digital
Front
End
digital analog
Base StationL Layers
July 5, 2019 | Blue Danube Systems | EXTERNAL | 3
R Digital Radio
Digital
ProcessorP
Regular
MIMO
P
R
...R
R
up to 8 radios
Digital
Full-Array
Massive
MIMO
P
R R R R R R
R R R R R R
R R R R R R
......
......
......
. . .. . .
. . .
128 radios
Digital
Sub-Array
Massive
MIMO
P
R R R
R R R
R R R
......
......
......
. . .. . .
. . .
32 radios
Regular MIMO and Digital Massive MIMO
OEM Systems
Hybridization to Lower Cost
Brute Force Scaling
July 5, 2019 | Blue Danube Systems | EXTERNAL | 4
Digital Massive MIMO Concept
• P=N=M=T = Large number such as 64, 128, 256 …
• Proposed for TDD by Marzetta/Ashikhmin in 2010.
• The main goal is to make MU-MIMO feasible
• All antenna signals are digitized and processed in the digital domain
• “Digital” Massive MIMO has massive amount of analog/RF circuits
and it is very expensive!
T
Antennas
L
LayersBaseband
Processor
N
ADC/DAC
Stages
M
RF
StagesP Ports
…
…
… …
UE1
UE2
UE3
Digital
Front
End
digital analog
Base StationL Layers
July 5, 2019 | Blue Danube Systems | EXTERNAL | 5
Motivation for Digital Massive MIMO
4 MU-MIMO Layers
Superimposed in RF Domain
4 MU-MIMO Layers
Separated In Digital Domain
(transmitted at RF)
(generated in digital domain)
▪ DSP-Based 3D Beamforming (BF)
▪ DSP-Based Multi-User MIMO (MU-MIMO)
both require
channel reciprocity!
July 5, 2019 | Blue Danube Systems | EXTERNAL | 6
Channel Sounding in TDD
UE
MIMO channel
including
the radio chains
.. .
received TDD pilots
measuring the channel
.. .
transmitted TDD pilots
aligned in phase
Downlink peaks (“beams”) and nulls (zeros, notches)
by Zero Forcing Algorithm
July 5, 2019 | Blue Danube Systems | EXTERNAL | 7
TDD Zero Forcing: Ideal Beam Placing
peak
Peak DSP-Based Beam
Layer 1 Layer 2
Layer 3 Layer 4
July 5, 2019 | Blue Danube Systems | EXTERNAL | 8
TDD Zero Forcing: Ideal Null Placing
zero zero
zero
Zero Notch Very Narrow Bandpass Filter
Layer 1 Layer 2
Layer 3 Layer 4
July 5, 2019 | Blue Danube Systems | EXTERNAL | 9
What’s in a Name?
That which we call a rose By any other name would smell as sweet!
(Romeo & Juliet)
Applying
Shakespeare’s genius
to Massive MIMO
An analog process By any other name (e.g. “Digital”) would be as tricky!
July 5, 2019 | Blue Danube Systems | EXTERNAL | 10
DSP-Based BF & MU-MIMO: Mostly Analog Schemes
DSPRadio
Array
UE
Channel + Transceiver Reciprocity
Channel Sounding
eNodeB
Analog signals & processing easily corrupted by noise and interference!
Digital signals & processing subject to “Garbage-In-Garbage-Out”
DSP Cannot Fix Major Errors in the Analog Portion of this Scheme!
July 5, 2019 | Blue Danube Systems | EXTERNAL | 11
Fundamental Flaws in Digital Massive MIMO
▪ Conventional Massive MIMO is designed for DSP-Based 3D BF & MU-MIMO
▪ DSP-Based techniques rely heavily on analog methods
❑ Channel sounding is an analog process
❑ Radio transceiver reciprocity is an analog design requirement
❑ Large # of radio chains yields a high content of analog & RF circuits
▪ What is digital in “Digital” Massive MIMO?
❑ Only digital computations for DSP-Based 3D beamforming
❑ Digital processing is subject to “Garbage-In-Garbage-Out”
▪ DSP-Based 3D Beamforming is highly vulnerable to inaccuracies in channel sounding and transceiver reciprocity
❑ Interference cancellation by Zero-Forcing algorithms is easily voided by real-world inaccuracies due to: mobility, interference & HW impairments.
July 5, 2019 | Blue Danube Systems | EXTERNAL | 12
The RF Coherency Difference - 2 GHz Band Example
0° RMS (0ps)
RF simulations with HFSS:
• 12x4 active array
• 0.5 λ (wavelength) horizontal spacing
• 0.7 λ vertical spacing
• Zero magnitude errors in all cases
150° RMS (204ps)
75° RMS (104ps)
25° RMS (35ps)
5° RMS (7ps) Necessary 3GPP HW Spec
July 5, 2019 | Blue Danube Systems | EXTERNAL | 13
TDD Zero Forcing Under Impairments
Layer 1 Layer 2
Layer 3 Layer 4
July 5, 2019 | Blue Danube Systems | EXTERNAL | 14
TDD Zero Forcing Under Impairments
Layer 1 Layer 2
Layer 3 Layer 4
July 5, 2019 | Blue Danube Systems | EXTERNAL | 15
TDD Zero Forcing Under Impairments
Layer 1 Layer 2
Layer 3 Layer 4
July 5, 2019 | Blue Danube Systems | EXTERNAL | 16
TDD Zero Forcing Under Impairments
Layer 1 Layer 2
Layer 3 Layer 4
July 5, 2019 | Blue Danube Systems | EXTERNAL | 17
Tx/Rx Reciprocity Challenge in TDD Digital Massive MIMO
DACTx Chanel
FilterPA
Rx Chanel
FilterADC LNA
Tx phase shift & gain
Rx phase shift & gain
Tx/Rx Reciprocity conditions• Tx phase shift = Rx phase shift
• Tx gain = Rx gain
Tx/Rx Reciprocity difficult• PA/LNA very different circuits
• Tx/Rx filters different circuits
• ADC/DAC different circuits
• Tx/Rx do not track with T, etc.
• Local T changes rapidly due to
PA activity
Reciprocity errors in TDD equivalent to RF Coherency errors in FDD
July 5, 2019 | Blue Danube Systems | EXTERNAL | 18
Typical Tx/Rx Variations with T◦ (Measured Data)
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
0.8
35 40 45 50 55
RF Front-End Gain
0
5
10
15
20
25
35 40 45 50 55
RF Front-End Phase
Temp [°C] Temp [°C]
Phase [°]
Mag [d
B]
Tx
Rx
Tx
Rx
Tem
p [°C
]
Time = one month
Measured Temperature Across 12x4 Panel
Important Lesson: Aligning all Tx and all Rx within 5◦
rms is excellent for TDD Rx/Tx
reciprocity!
70
60
50
40
30
July 5, 2019 | Blue Danube Systems | EXTERNAL | 19
The Holy Grail of Wireless: True 3D Beams
What are True 3D Beams?
• Narrow EM radiation patterns also known as “Pencil Beams”
• Most radiated EM energy focused into a cone with little EM energy outside this cone
• Best use of EM spectrum (best SNR) for carrying information between two points
• Least amount of interference to others: Best for spectrum reuse!
• 5G targets extensive use of True 3D Beams
July 5, 2019 | Blue Danube Systems | EXTERNAL | 20
MU-MIMO by True 3D Beams
4 Layers Superimposed4 Layers Separated
(transmitted at RF)
(generated in digital domain)
July 5, 2019 | Blue Danube Systems | EXTERNAL | 21
Not All Beamforming Methods Are Created Equal!
DSP-Based Beamforming True 3D Beams
July 5, 2019 | Blue Danube Systems | EXTERNAL | 22
How Can We Generate True 3D Beams?
Dish Antenna Phased Array
▪ Beamforming by dish geometry
▪ Beam direction by mechanical
orientation
▪ Low cost but bulky, single beam,
not agile and not SW programable
▪ Beamforming by RF Coherency
▪ Beam direction by electrical phasing
(no mechanical movement)
▪ Very expensive but agile, multiple
beams and SW programmable
July 5, 2019 | Blue Danube Systems | EXTERNAL | 23
Classical Planar
Phased Array*
* www.radartutorial.eu
High Definition Active Antenna
System™ (HDAAS™)Commercial Deployment
of HDAAS™
Phase Array Cost
Breakthrough
by Blue Danube
Application
to 4G & 5G
First FDD Massive MIMO
October 2016
$$$$ $
Exclusive Low-Cost Phased Arrays in LTE Operation!
July 5, 2019 | Blue Danube Systems | EXTERNAL | 24
Blue Danube RF Coherency at 2 GHz
0° RMS (0ps)
RF simulations with HFSS:
• 12x4 active array
• 0.5 λ (wavelength) horizontal spacing
• 0.7 λ vertical spacing
• Zero magnitude errors in all cases
150° RMS (204ps)
75° RMS (104ps)
25° RMS (35ps)
5° RMS (7ps)
Phased Array Quality
Blue Danube 3GPP Spec
July 5, 2019 | Blue Danube Systems | EXTERNAL | 25
Shentel, US
Single site trial
24+ months, 2X gain
AT&T, US
Urban/Residential environment
100% capacity increase
Telstra, Australia
High traffic coverage
5x user throughput
Shentel, US
Multi-site trial
2.5X capacity gain,
time of day beam
adjustment
Phase-1: 2.1X capacity on 3 sector site
Phase-2: Adding panel on facing site
2016
2017
2018
Large Asian Operator
> 90% utilized sites
2.7X capacity gain,
3X traffic off-load,
vertical beam gainBeamCraft™ Product Family
AT&T, Mexico, Dense urban
Proven Performance in Commercial Networks
July 5, 2019 | Blue Danube Systems | EXTERNAL | 26
Reducing the Complexity of Digital Massive MIMO
Add ICs with ph/mag ctr.
Different C&M
Same aperture
Same # Layers
Less MU-MIMO Complexity
(LxM) data
N analog (LxM) radios
N radios
Add precise RF calibration
Digital Massive MIMOHDAASTypical 4G LTE MIMO
July 5, 2019 | Blue Danube Systems | EXTERNAL | 27
4x12 Array Boresight: Azimuth & ElevationSimulation and RF Scanner Measured Data
Uniform Taylor Taper
July 5, 2019 | Blue Danube Systems | EXTERNAL | 28
4x12 Array: Taylor in Azimuth & Elevation
Far field measured data (major antenna OEM) and HFSS simulation overlays
July 5, 2019 | Blue Danube Systems | EXTERNAL | 29
RF Energy
optimally
placed
Blue Danube
RF Placement Cloud SW
Blue Danube Smart RF Placement & Virtualization
ADC/DAC
Bank
RF
StagesPorts
…
…
… …
UE1
UE2
UE20Digital
Front
End
UE21
UE45
SON
EPC
BBUnon-uniform
trafficBlue Danube Hybrid Massive MIMO
▪ The baseline radiation pattern of each radio chain can be set
from the Cloud using ML and AI methods!
▪ All BBU processing is same as for any other MIMO system,
including Massive MIMO
TMx
July 5, 2019 | Blue Danube Systems | EXTERNAL | 30
Examples of Cloud-Programable Beam Patterns
Wide Beam
Narrow
Beam
Amoeba Beam V-Beam
July 5, 2019 | Blue Danube Systems | EXTERNAL | 31
Coherent Massive MIMO Upgrade Roadmap
Standard
4G BBU
Standard SON + Adaptive 3D beams matching traffic & clutter
Standard 4G/5G BBU/DU
with MU-MIMO
or Massive-MIMO DU
SW Upgrade SW Upgrade
Standard
5G DU
4G True 3D Beams
Flexible Sectorization
5G True 3D Beams
Flexible Sectorization
CPRI VRAN CPRI/VRAN
4G/5G MU-MIMO
with True 3D Beams
Upgrade Upgrade
same HDAAS on tower
Tower
Ground
Cloud
July 5, 2019 | Blue Danube Systems | EXTERNAL | 32
▪ Produces predictable True 3D Beams without channel sounding & DSP
❑ Avoids the fundamental flaws of standard Massive MIMO
✓ Far less performance degradation to real-world conditions: low SNIR, mobility, cluster deployment, number of users served
✓ Much better performance in FDD bands (lack of channel reciprocity)✓ Easy to generate vertical narrow beams for high-rises
❑ Best for MU-MIMO
✓ UE separation at RF before MU-MIMO processing
❑ Significant reduction in HW compared to standard Massive MIMO
✓ Lower cost, lower weight and lower power dissipation for FDD and TDD
▪ Enables easy cloud interference management: Real-Time RF placement
▪ Enables digital sub-band beamforming for hybrid architectures
RF Coherency Brings in Many Advantages
July 5, 2019 | Blue Danube Systems | EXTERNAL | 33
Four Key Parameters of MIMO Systems
1. Number of passive antennas
❑ Defines the array aperture, directly responsible for narrowest 3D beam possible
2. The resolution of the active array
❑ Number of subarrays that can be controlled independently in phase and magnitude
❑ The higher the resolution the smaller the side lobes
3. The number of digitizers in the active array
❑ “Digitizer” used here as synonymous term to “radio”, “digital radio” or “radio chain”
❑ Digitizers are expensive!
4. The number of data streams (Layers) that may be transmitted in parallel
❑ Layer is 3GPP terminology
❑ The more Layers the higher system capacity
R
...
. . .. . .
......
...
...
R
...
. . .. . .
......
...
...
. . .. . .
......
...
...
. . .. . .
......
...R
...
R
beam width
main beam
radio chains
Data 1
Data 2
Expensive
phase/mag
control
APERTURE
RESOLUTION
DIGITIZATION
LAYERS
July 5, 2019 | Blue Danube Systems | EXTERNAL | 34
Evolution of MIMO Systems for Cellular
Common
4G LTE MIMO
(Release 8-11)
Academic Research
“Marzetta”
Massive MIMO
Coherent
Massive MIMO
mmWave & sub 6 GHz
Sub 6 GHz
Digital Sub-Array
Massive MIMO
fixed
antenna
mapping
fixed
antenna
mappingfixed
sub-array
mapping dynamic
array
mapping
scalable
4G 5G
Blue Danube
July 5, 2019 | Blue Danube Systems | EXTERNAL | 35
Conclusions
1. Conventional Digital Massive MIMO has fundamental flaws
2. Array RF coherency avoids the fundamental flaws of Conventional Digital Massive MIMO
3. Digital signal processing alone cannot produce array RF coherency
4. Economical FDD Coherent Massive MIMO has been demonstrated
July 5, 2019 | Blue Danube Systems | EXTERNAL | 36
Thank You